Potential clipping and tiling issue with focalCpp for large rasters on Linux

[ailich]

I believe there may be an issue occurring in terra::focalCpp for data too large to fit in memory (potentially specific to Linux systems). While I get the correct result on a Windows computer with 16GB of RAM and another user is getting correct results on a Linux computer with 128GB of RAM, their Linux computer with 16 GB of RAM is providing clipped/offset/striped outputs from the call to focalCpp and on an Ubuntu VM with ~10GB of RAM I'm getting clipped/offset results on two different functions that use focalCpp internally. When I overlay the original data with a single layer of the output you can see the two do not match up. Additionally, on Windows focalCpp for the second example results in a memory exhausted error. For more details see ailich/GLCMTextures#12. The data set is available here .

GLCMTextures Example

Using the function glcm_textures from the GLCMTextures package.

Overlay of the original data and a single layer of intermediate surface calculated with focalCpp plotted from debug mode on 10GB Linux VM

Intermediate surface calculated with focalCpp plotted from debug mode on 16GB Linux Computer

Surfaces from a 16GB Windows Computer (correct)

MultiscaleDEM Examples

Using the function Qfit from the MultiscaleDEM package

Overlay of a single layer of the output onto the original data from the 10GB Linux VM

Windows 16GB Computer
Error: memory exhausted (limit reached?)

[rhijmans]

Closing because this is the same as #519 (fixed but awaiting confirmation)